Betterleaks Cloud - GitHub & S3 Secret Scanner

Cloud-hosted Betterleaks v1.3.1 for hunting leaked API keys, wallet private keys, and credentials anywhere on GitHub, in cloned git repos, or in S3 / R2 / MinIO buckets. Optional live validation probes each detected secret against the vendor API to flag which ones are still active.

Available as an Apify Actor. $0.05 start + $0.05 per repo scanned + $0.30 per live-validated key. No install, no CLI, flat JSON output.

What does it do?

Scans a GitHub org, user, repo, pull request, issue, gist, action workflow, or S3 bucket for leaked credentials. Built on Betterleaks (the Gitleaks successor, maintained by the original Gitleaks author Zachary Rice + co-maintainers from Red Hat, Amazon, RBC). Every Betterleaks CLI flag is exposed as an input field. When you enable validation, each finding is probed live against the vendor API and tagged valid, invalid, revoked, unknown, or error.

How is it different from running Betterleaks locally?

How is it different from our gitleaks-github-secret-scanner actor?

When should I use it?

• Auditing your own GitHub org for secrets that slipped into commits
• Bug bounty hunting where you want to validate keys live before reporting
• Pre-acquisition security check on a target company's GitHub footprint
• Continuous monitoring on a schedule, alerting on any new live-validated finding
• Forensics: tracing which dangling commit on which PR ref leaked a credential
• Scanning a Cloudflare R2 / AWS S3 bucket someone shared publicly

What does it cost?

Pay-per-event:

• $0.05 per actor start
• $0.05 per repo scanned (in github / git / global_github modes)
• $0.01 per finding pushed to the dataset
• $0.30 per live-validated secret (only when validation is on AND status returns valid)
• $0.005 per Code Search query (only in global_github mode)

Typical scans:

Compare to GitGuardian Enterprise at ~$5000/year flat. Per-comprehensive-scan with this actor is 300-500x cheaper.

Which modes can I scan in?

Which GitHub resources can it scan in github mode?

All 12 surfaces betterleaks supports, selectable via include / exclude:

• repos (default), forks
• prs, pr-comments
• issues, issue-comments
• actions, action-artifacts
• discussions
• releases, release-assets
• gists

You can also filter by since / until dates, exclude repos by glob pattern, or target specific GitHub Actions workflow files.

Does it find leaks that were deleted from current files?

Yes. Betterleaks scans the full git history. A secret that was committed and then deleted in a later commit still appears in the history and gets detected. The Commit field on each finding tells you which commit introduced the leak.

You can also enable advanced --log-opts pass-through to use git pickaxe (-S secret_string) for targeted historical searches.

Is the secret validated against the live vendor API?

Optional - off by default to keep scans fast. Set validation: true and Betterleaks fires HTTP requests against the vendor API for each finding using the rule's built-in CEL validator.

Output fields:

• ValidationStatus: valid (still active), invalid (rejected), revoked (explicitly disabled), unknown (couldn't determine), or error
• ValidationReason: human-readable why
• ValidationMeta: vendor-specific metadata. For a GitHub PAT, this includes username, name, and granted scopes. For Stripe, the test/live mode + account info.

Note: validation only fires for rules that ship with a validate CEL clause in their TOML. Not every detected secret type has one (the upstream rule set is expanding over time).

You can also tune validation behavior:

• validation_timeout (per-request HTTP timeout)
• validation_workers (parallel HTTP workers)
• validation_status (filter output to only certain statuses)
• validation_debug (include raw HTTP response in ValidationMeta)
• validation_extract_empty (include validator outputs even when empty)
• validation_env_vars (expose extra env vars to CEL validators)

Can I use a custom detection rule?

Yes - paste a full betterleaks TOML config into custom_config_toml. Either replaces the built-in rules or adds to them depending on how you write the TOML.

You can also use enable_rule to run only a whitelist of rule IDs (e.g. ["github-fine-grained-pat", "stripe-secret-key"]) from the built-in set without writing your own config.

Reference format: https://github.com/betterleaks/betterleaks/blob/main/.betterleaks.toml

What does the output look like?

Each dataset record is the raw Finding struct from betterleaks - no transformation, no field renaming:

{
  "RuleID": "github-fine-grained-pat",
  "Description": "GitHub Fine-Grained Personal Access Token, risking unauthorized repo access.",
  "Match": "github_pat_11AABBCC...",
  "Secret": "github_pat_11AABBCC...",
  "StartLine": 12,
  "EndLine": 12,
  "StartColumn": 18,
  "EndColumn": 105,
  "MatchContext": "...optional surrounding lines if match_context is set...",
  "CaptureGroups": {},
  "Fragment": {
    "FilePath": "config/secrets.env",
    "Url": "https://github.com/owner/repo/blob/<sha>/config/secrets.env#L12"
  },
  "Attributes": {
    "path": "config/secrets.env",
    "resource": "git.patch_content",
    "url": "https://github.com/owner/repo/blob/<sha>/config/secrets.env#L12",
    "git.author_name": "alice",
    "git.author_email": "alice@example.com",
    "git.commit": "abcd1234",
    "git.date": "2025-06-12T10:23:14Z",
    "git.message": "Add config",
    "github.owner": "owner",
    "github.repo": "repo",
    "github.visibility": "public"
  },
  "Tags": [],
  "Fingerprint": "abcd1234:config/secrets.env:github-fine-grained-pat:12",
  "ValidationStatus": "valid",
  "ValidationReason": "",
  "ValidationMeta": {
    "username": "alice",
    "name": "Alice Doe",
    "scopes": "repo, workflow"
  }
}

The dataset has two pre-configured views:

• Findings: all records
• Live secrets only: filter to ValidationStatus == valid, the high-signal triage list

How accurate is it?

Based on the underlying Betterleaks 1.3.1 rule set (200+ detectors), plus the upstream's BPE-tokenization based false-positive filter and CEL-based contextual filters. Compute usage observed in real test runs:

• Single small repo scan: ~7 seconds, 0.0084 compute units
• Scan of anshumanatrey/* (~30 repos): ~134 seconds, 0.1491 compute units, 735 findings
• 3-repo global_github scan with query rzp_live_: ~8.5 seconds, 0.0095 compute units, 14 findings

Common questions

Q: Can I scan a private repo? Yes - provide a github_token with repo scope.

Q: Does it scan PR branches and dangling commits? PR refs yes via --include prs. Dangling commits aren't currently exposed but the underlying tool supports it.

Q: How does it handle rate limits? Without a github_token, GitHub limits to ~60 req/hr shared. With a token: 5000 req/hr on your account. Always provide a token for any real scan.

Q: Can I run it on a schedule? Yes via Apify Schedules. Recommended for continuous monitoring use cases.

Q: Can I use my own betterleaks rule TOML? Yes via custom_config_toml. Or whitelist specific built-in rules via enable_rule.

Q: What's global_github mode? Our addition on top of upstream betterleaks. You provide a search query, we use GitHub Code Search to find unique candidate repos, then run betterleaks git against each in parallel. Useful for hunting a specific secret pattern across all of GitHub.

Q: Why not use the cheaper gitleaks-github-secret-scanner actor instead? Use that one for high-volume cost-sensitive scans. Use this one when live validation matters (bug bounty triage, incident response, knowing which secrets to actually rotate).